4  Intermediate Git Workflow

The intermediate Git workflow introduces a new concept: branches. Efficient use of branches and a well thought out branching strategy will aid the analyst in Git use and is also a powerful tool for model development.

4.0.1 Branches and Branch strategy

In Git a branch is a pointer to a specific commit or set of commits which allow you to separate model development tasks into smaller subunits. I learned branching from this guy and in what follows I simplify his workflow into something that works well for complicated fisheries analyses. Let’s start by creating a new branch named develop using the command git checkout. The argument -b tells Git to checkout a new branch while the arguments develop and main tell Git the name of the new branch and that the new branch should branch from the main branch.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (main)
$ git checkout -b develop main
Switched to a new branch 'develop'

The core concept in the Git Flow branching strategy is to always have two active branches; main and develop. The main branch is stable in that you make limited commits to it and those commits are associated with internal or external reporting including FDS reports, BOF memos, or conference presentations. Because the main branch will largely be static interested collaborators can quickly identify the analysis at the time periods where the analysis was reported.

The develop branch is the working branch and will have frequent commits relative to analysis progress. The develop branch merges into main at reporting periods and is the branched from whenever a new feature is being developed.

Feature branches are created frequently as new features are envisioned and developed. A best practice is to create a new feature branch, with a descriptive name, every time your create a new feature. When the feature is completed it is merged back into develop, left as a record without a merge, or deleted. In my work, feature branches are mostly merged back into develop, and this occurs whenever I create a feature which improves the analysis. Feature branches are left unmerged when I want to retain a record of having tried something (and the result) but do not think it improves the overall analysis. Feature branches are deleted without merging when something just did not work out and is also not worth retaining as a record. Notice that liberal use of feature branches keep the main and develop branches clean and can isolate those two primary branches of a lot of the sloppiness that is a byproduct of actively engaging in the scientific process.

Let’s practice Git Flow. In the terminal session which follows we create a new branch called cleanup with git checkout, modify the file fib_seq.R by deleting several lines, show the changes using git diff, add those changes, and commit those changes.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git checkout -b cleanup develop
Switched to a new branch 'cleanup'

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (cleanup)
$ git status
On branch cleanup
Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
        modified:   fib_seq.R

no changes added to commit (use "git add" and/or "git commit -a")

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (cleanup)
$ git diff
diff --git a/fib_seq.R b/fib_seq.R
index 984a0a5..97be5b3 100644
--- a/fib_seq.R
+++ b/fib_seq.R
@@ -2,12 +2,7 @@
 #Adam Reimer & Carly Reimer

 fib_seq <- c(0, 1)
-fib_seq[3] <- fib_seq[1] + fib_seq[2]
-fib_seq[4] <- fib_seq[2] + fib_seq[3]
-fib_seq[5] <- fib_seq[3] + fib_seq[4]
-fib_seq[6] <- fib_seq[4] + fib_seq[5]
-fib_seq[7] <- fib_seq[5] + fib_seq[6]
-for (i in 8:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])
+for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

 plot(1:50, fib_seq[2:51]/fib_seq[1:50], type = "l")
-{golden_ration <- fib_seq[51]/fib_seq[50]}
+{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (cleanup)
$ git add fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (cleanup)
$ git commit -m "remove maanual Fibonacci calculations" -m "manual calculations become redundent now that we are using recurstion to get the series."
[cleanup 8b28369] remove maanual Fibonacci calculations
 1 file changed, 2 insertions(+), 7 deletions(-)

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

4.0.2 git merge

The changes above are reasonable and the new feature tried on the cleanup branch is useful. In the terminal session below we use git checkout to switch back to the develop branch, git merge to merge the cleanup branch with the develop branch, and git branch to delete the cleanup branch now that it’s change has been incorporated into develop.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (cleanup)
$ git checkout develop
Switched to branch 'develop'

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git merge --no-ff cleanup
Merge made by the 'ort' strategy.
 fib_seq.R | 9 ++-------
 1 file changed, 2 insertions(+), 7 deletions(-)

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git branch -d cleanup
Deleted branch cleanup (was 8b28369).

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

4.0.2.1 Merge conflicts

Merge conflicts can occur any time we are merging two branches. In what follows we will intentionally create a merge conflict and fix it. In the terminal session below I create a new branch named label_plot, and while working on that branch modify the file fib_seq.R by adding informative figure labels and then stage/commit those changes.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git checkout -b label_plot develop
Switched to a new branch 'label_plot'

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (label_plot)
$ git status
On branch label_plot
Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
        modified:   fib_seq.R

no changes added to commit (use "git add" and/or "git commit -a")

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (label_plot)
$ git add fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (label_plot)
$ git commit -m "Added axis labels to golden ratio plot." -m "Not a lot to say but hammering home the idea that more informative messages are better. In this case I added labels to demonstrate a merge conflict. Will add different labels to the same file/figure on branch develop to create the conflict."
[label_plot a4c0f61] Added axis labels to golden ratio plot.
 1 file changed, 4 insertions(+), 1 deletion(-)
 
amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (label_plot)
$

A merge into the develop branch at this stage would be both appropriate and successful. The new feature was added to the feature branch label_plot was a improvement, and the develop branch is unchanged. Instead we will intentionally change the develop branch (violating the spirit of Git Flow) to demonstrate a merge conflict. In the terminal session that follows we: checkout the develop branch and while working on that branch modified the file fib_seq.R by adding less informative figure labels, and stage/commit those changes. We then use git show to look at the copy of fib_seq.R from both the develop and label_plot branches. Notice that the axis labels from the label_plot branch are more informative. Because the values given to the xlab and ylab arguments were modified in both branches we can expect a merge conflict and, as expected, a merge conflict results from attempting to merge these branches.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (label_plot)
$ git checkout develop
Switched to branch 'develop'

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
        modified:   fib_seq.R

no changes added to commit (use "git add" and/or "git commit -a")

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git add fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git commit -m "Added labels to the golden ratio figure" -m "These labels are intentionally less informative that the labels added on the feature branch to demonstrate that feature branches are where you do the work which you then merge into develop."
[develop d0d260e] Added labels to the golden ratio figure
 1 file changed, 4 insertions(+), 1 deletion(-)

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git show develop:fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Number",
     ylab = "Ratio")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git show label_plot:fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git merge --no-ff label_plot
Auto-merging fib_seq.R
CONFLICT (content): Merge conflict in fib_seq.R
Automatic merge failed; fix conflicts and then commit the result.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

When there is a merge conflict Git modifies the file being merged to help the user decide how to proceed. Git adds <<<<<<< HEAD above the merge conflict. The text ====== separates the merge conflict into code derived from each branch. Code between <<<<<< HEAD and ====== indicates code which was present in the branch being merged into (develop in this case) while code between ====== and >>>>>> label_plot indicates code which was present in the branch being merged (label_plot in this case). I’ve shown what that fib_seq.R file in my working directory looks like using the terminal command cat, but that is only for demonstrative purposes. Normally, I would just open the file in RStudio, and it would look exactly the same. To proceed the user has to decide which code to keep, save the modified file, and stage/commit the changes. For this example I deleted everything between <<<<<< HEAD and ====== inclusive as well as the line >>>>>> label_plot. These deletions made the file fib_seq.R we merged match the version in the label_plot branch.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop|MERGING)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
<<<<<< HEAD
     type = "l",
     xlab = "Number",
     ylab = "Ratio")
======
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
>>>>>> label_plot
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop|MERGING)
$ git status
On branch develop
You have unmerged paths.
  (fix conflicts and run "git commit")
  (use "git merge --abort" to abort the merge)

Unmerged paths:
  (use "git add <file>..." to mark resolution)
        both modified:   fib_seq.R

no changes added to commit (use "git add" and/or "git commit -a")

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop|MERGING)
$ git add fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop|MERGING)
$ git commit -m "fixed merge conflict by accepting changes from label_plot branch" -m "Mmaybe the only good example of a situation were a description is not nessesary."
[develop ad2149a] fixed merge conflict by accepting changes from label_plot branch

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
nothing to commit, working tree clean

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

4.0.3 Interacting with your Git history

4.0.3.1 Restoring a previous version of a file

You can also use git checkout to recall a single file from a previous commit. In the terminal sequence below I use cat and git show to show that the file in my work directory and the file in the most recent commit to the develop branch are the same file. Notice a call to git status reports the same thing without showing the files explicitly.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git show develop:fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
nothing to commit, working tree clean

To retrieve a specific file from a prior commit you have to provide git checkout with identifying information for the commit and the file name. You can specify the appropriate commit with a absolute or relative reference. For an absolute reference provide the SHA #. For a relative reference the shorthand is HEAD~# where # is the number of commits prior to the latest commit. In the terminal sequence below I used checkout to retrieve a copy of fib_seq.R from 1 commit earlier (recall this commit had the less informative x and y axis labels). Notice the earlier version of fib_seq.R was present in my working directory (as shown in the output for cat) but was not included in the latest commit to my repository (as shown in the output to git show). If I wanted the earlier version of fib_seq.R to be associated with the most recent commit to my repository I would need to commit the file. In this case I did not want to change my repository so I used git checkout to go back the the most recent version. I often use this sequence to recall all or part of a code from a prior commit.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git checkout HEAD~1 fib_seq.R
Updated 1 path from abb8fba

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Number",
     ylab = "Ratio")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git show develop:fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
Changes to be committed:
  (use "git restore --staged <file>..." to unstage)
        modified:   fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git checkout HEAD fib_seq.R
Updated 1 path from bd65128

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

4.0.3.2 Retrieving a prior commit

Yet another use of git checkout is to recall a previous commit in it’s entirety. Since our repository only has one file I created, staged and committed a second file named test.R to demonstrate this functionality. In the terminal sequence below I show the contents of test.R and some now familiar content where I check the status of my repository and stage/commit the new file.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat test.R
#Blank file to see if we can use git checkout to retrieve an entire commit.
#Proof would be seeing this file and the correct version of fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
Untracked files:
  (use "git add <file>..." to include in what will be committed)
        test.R

nothing added to commit but untracked files present (use "git add" to track)

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git add test.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git commit -m "Test.R file added. Used to test that git checkout recalls an entire commit" -m "added a second file so to verify both files come when I jump around commits."
[develop 4ef90af] Test.R file added. Used to test that git checkout recalls an entire commit
 1 file changed, 2 insertions(+)
 create mode 100644 test.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

To retrieve a prior commit you should create a new branch to house the past commit until you decide what to do with it. While not strictly necessary creating a new branch is stable and protects your develop branch. In the terminal session below I use git checkout to create a new branch called old_state and fill it with the snapshot of the develop branch from 2 commits earlier. I am expecting old state to be missing test.R and have the less descriptive axis locations in the version of fib_seq.R. Two calls to cat confirm that my working directory does not contain test.R and has the older version of fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git checkout -b old_state HEAD~2
Switched to a new branch 'old_state'

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (old_state)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Number",
     ylab = "Ratio")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (old_state)
$ cat test.R
cat: test.R: No such file or directory

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (old_state)
$ git status
On branch old_state
nothing to commit, working tree clean

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

The next action depends on if you want the keep working on the prior commit. If you do you can modify files and stage/commit the result as appropriate prior to merging old_state back into develop. Since you purposely started the old_state feature branch from a prior version of develop the odds of a merge conflict are high. It does mean it’s a terrible idea but you should be doing this with intention. In the terminal session below I decided the prior commit was not an improvement on my original commit so I moved back to the develop branch and deleted the old_state branch. Two calls to cat prove that the original files were returned.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (old_state)
$ git checkout develop
Switched to branch 'develop'

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat test.R
#Blank file to see if we can use git checkout to retrieve an entire commit.
#Proof would be seeing this file and the correct version of fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git branch -d old_state
Deleted branch old_state (was d0d260e).

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

4.0.3.3 Deleating/Reversing a commit

There are two ways to reverse a commit. The version you should choose depends on if you have pushed the commit to a remote repository.

4.0.3.3.1 Deleting a local commit

When the bad commit is local you can us git reset to delete the commit. The commit will disappear from your history. To demonstrate this command I modified the fib_seq.R by adding an inconsequential comment at the bottom of the file and used the cat command to show the file. After adding and committing the file a call to git log shows that the commit was recorded.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

#Useless comment to demonstrate how to delete a commit

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git add fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git commit -m "bad commit, plan to delete" -m "This commit has one change, a useless comment added to the fib_seq.R file. Plan to reset the HEAD to delete the commit"
[develop ae2fe59] bad commit, plan to delete
 1 file changed, 2 insertions(+)

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git log --oneline
ae2fe59 (HEAD -> develop) bad commit, plan to delete
4ef90af Test.R file added. Used to test that git checkout recalls an entire commit
ad2149a fixed merge conflict by accepting changes from label_plot branch
d0d260e Added labels to the golden ratio figure
a4c0f61 Added axis labels to golden ratio plot.
eb216a6 Merge branch 'cleanup' into develop
8b28369 remove maanual Fibonacci calculations
22dcfea (origin/main, main) Merge pull request #2 from carlyreimer/main
f811c50 Update fib_seq.R
edbaafa Update fib_seq.R
ac48c12 Finally added a for loop and golden ratio estimate
f732cdb Seventh Fibonacci number
9db5478 Sixth number in the Fibonacci seqence
5139049 Fifth entry in the Fibonacci sequence
0c92881 Fourth entry in the Fibonacci sequence
3bb6c98 Third entry in fib_seq
e17181f Initialize Fibonacci sequence
    
amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$    

In the terminal session below I use git reset to reset my local repository to 1 commit prior to the most recent commit. Notice that the comment I added to the most recent commit is no longer a part of the fib_seq.R file and that the commit ae2fe59d44eec001b949c3677f94f78e0c30a216 has been removed for the repository. The important thing to note is git reset changes the repository history. If you are working locally this is OK but if you have already pushed the bad commit to a remote then deleting a part of your history will cause problems for collaborators who pulled the original history.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git reset HEAD~1 --hard
HEAD is now at 4ef90af Test.R file added. Used to test that git checkout recalls an entire commit

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ cat fib_seq.R
#Create a Fibonacci sequence to practice git operations
#Adam Reimer & Carly Reimer

fib_seq <- c(0, 1)
for (i in 3:51) fib_seq[i] <- sum(fib_seq[(i-1):(i-2)])

plot(1:50, fib_seq[2:51]/fib_seq[1:50],
     type = "l",
     xlab = "Fibonacci Number in Denominator",
     ylab = "Ratio Between Succesive Fibonacci Numbers")
{golden_ratio <- fib_seq[51]/fib_seq[50]}

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git log --oneline
4ef90af (HEAD -> develop) Test.R file added. Used to test that git checkout recalls an entire commit
ad2149a fixed merge conflict by accepting changes from label_plot branch
d0d260e Added labels to the golden ratio figure
a4c0f61 Added axis labels to golden ratio plot.
eb216a6 Merge branch 'cleanup' into develop
8b28369 remove maanual Fibonacci calculations
22dcfea (origin/main, main) Merge pull request #2 from carlyreimer/main
f811c50 Update fib_seq.R
edbaafa Update fib_seq.R
ac48c12 Finally added a for loop and golden ratio estimate
f732cdb Seventh Fibonacci number
9db5478 Sixth number in the Fibonacci seqence
5139049 Fifth entry in the Fibonacci sequence
0c92881 Fourth entry in the Fibonacci sequence
3bb6c98 Third entry in fib_seq
e17181f Initialize Fibonacci sequence

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
nothing to commit, working tree clean

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

4.0.3.3.2 Reversing a local/remote commit

Because revising your history in a public repository is bad collaboration Git has devised a solution. The git revert command can also be used to undo a mistaken commit but it does so in a very deliberate way… by creating a new commit that exactly undoes the prior commit. This method provides a stable Git history to your collaborators. In the git session below I modified the fib_seq.R again (with an inconsequential comment) and added/committed the change. git log shows the new commit.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
        modified:   fib_seq.R

no changes added to commit (use "git add" and/or "git commit -a")

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git add fib_seq.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git commit -m "bad commit, plan to revert" -m "This commit has one change, a useless comment added to the fib_seq.R file. Plan to revert the commit""
> "
[develop a0c849e] bad commit, plan to revert
 1 file changed, 2 insertions(+)

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git log --oneline
a0c849e (HEAD -> develop) bad commit, plan to revert
4ef90af Test.R file added. Used to test that git checkout recalls an entire commit
ad2149a fixed merge conflict by accepting changes from label_plot branch
d0d260e Added labels to the golden ratio figure
a4c0f61 Added axis labels to golden ratio plot.
eb216a6 Merge branch 'cleanup' into develop
8b28369 remove maanual Fibonacci calculations
22dcfea (origin/main, main) Merge pull request #2 from carlyreimer/main
f811c50 Update fib_seq.R
edbaafa Update fib_seq.R
ac48c12 Finally added a for loop and golden ratio estimate
f732cdb Seventh Fibonacci number
9db5478 Sixth number in the Fibonacci seqence
5139049 Fifth entry in the Fibonacci sequence
0c92881 Fourth entry in the Fibonacci sequence
3bb6c98 Third entry in fib_seq
e17181f Initialize Fibonacci sequence

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

Using git revert I specify the short SHA id from the prior commit. The call to git log shows that a new commit was added with a message specifying the commit reverted the commit titled “bad commit, plan to revert”.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git revert a0c849e
[develop 6466f85] Revert commit titled: "bad commit, plan to revert"
 1 file changed, 2 deletions(-)

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git log --oneline
6466f85 (HEAD -> develop) Revert commit titled: "bad commit, plan to revert"
a0c849e bad commit, plan to revert
4ef90af Test.R file added. Used to test that git checkout recalls an entire commit
ad2149a fixed merge conflict by accepting changes from label_plot branch
d0d260e Added labels to the golden ratio figure
a4c0f61 Added axis labels to golden ratio plot.
eb216a6 Merge branch 'cleanup' into develop
8b28369 remove maanual Fibonacci calculations
22dcfea (origin/main, main) Merge pull request #2 from carlyreimer/main
f811c50 Update fib_seq.R
edbaafa Update fib_seq.R
ac48c12 Finally added a for loop and golden ratio estimate
f732cdb Seventh Fibonacci number
9db5478 Sixth number in the Fibonacci seqence
5139049 Fifth entry in the Fibonacci sequence
0c92881 Fourth entry in the Fibonacci sequence
3bb6c98 Third entry in fib_seq
e17181f Initialize Fibonacci sequence

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git status
On branch develop
nothing to commit, working tree clean

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$

4.0.4 Before you go!

I get it, everyone is tired and i have beat this subject to death. Still, we should tidy up before we close the git-practice repository for an extended amount of time. The repository may or may not have any enduring value but it helped me get through this book. As such the the repository might be about to go public. That being the case the Git Flow workflow suggests we: merge develop back into main, tag the published version, and push the published vesrion to the remote. Lets do that to close out.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (develop)
$ git checkout main
Switched to branch 'main'
Your branch is up to date with 'origin/main'.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (main)
$ git merge --no-ff develop
Merge made by the 'ort' strategy.
 fib_seq.R | 14 ++++++--------
 test.R    |  2 ++
 2 files changed, 8 insertions(+), 8 deletions(-)
 create mode 100644 test.R

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (main)
$ git tag -a "v1.1" -m "This tag is synchronized with the first draft of the git book."

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (main)
$ git tag
v1.1

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (main)
$ git show v1.1
tag v1.1
Tagger: Adam Reimer <adam.reimer@alaska.gov>
Date:   Tue Jul 11 15:14:24 2023 -0800

This tag is synchronized with the first draft of the git book.

commit ef33f10150d6248b61558f63f98bd1d5acb3d771 (HEAD -> main, tag: v1.1)
Merge: 22dcfea 6466f85
Author: Adam Reimer <adam.reimer@alaska.gov>
Date:   Tue Jul 11 15:08:02 2023 -0800

    Merge branch 'develop'


amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (main)
$ git push -u origin main
Enumerating objects: 21, done.
Counting objects: 100% (21/21), done.
Delta compression using up to 16 threads
Compressing objects: 100% (19/19), done.
Writing objects: 100% (19/19), 2.98 KiB | 55.00 KiB/s, done.
Total 19 (delta 6), reused 0 (delta 0), pack-reused 0
remote: Resolving deltas: 100% (6/6), done.
To https://github.com/adamreimer/git_practice.git
   22dcfea..ef33f10  main -> main
branch 'main' set up to track 'origin/main'.

amreimer@DFGSXQDSF206801 MINGW64 /s/RTS/Reimer/Research_Best_Practices/git_practice (main)
$